H3K4 trimethylation at Notch target gene Deltex-1 is dynamic only at the RBP-J-binding site. (A) ChIP experiments were performed using anti-H3K4me3 antibodies and the pre-T-cell line Beko treated as in Figure 1D. When Notch signaling is switched off by treatment with GSI, H3K4 trimethylation is lost specifically at the conserved RBP-J-binding site (−1.2-kb upstream, red asterisk) but not at the TSS. The sequence of the RBP-J-binding site information can be found in Supplemental Figure S10. H3K4 trimethylation is re-established at the RBP-J-binding site upon removal of GSI (24 h). Shown are the values as mean ± SD for duplicate samples from a representative experiment; the experiment was repeated three times. H3K4me3 is significantly enriched with P < 0.05 (Student's t-test) at the RBP-J-binding site (−24 h, 24 h) and at the TSS. (B) Primary CD4⁻/CD8⁻ precursor T cells from RAG1 knockout mice have a similar H3K4me3 profile at the Deltex-1 promoter compared with Beko T cells (shown in A); H3K4 trimethylation is observed at the RBP-J-binding (−1.2-kb upstream) site and at the TSS. H3K4me3 is significantly enriched with P < 0.05 (Student's t-test) at the RBP-J-binding site and at the TSS. (C) The nucleosome occupancy does not change before or after GSI treatment, as measured by ChIP with antibodies directed against the core of Histone H3. (D) Nucleosomal structure, indicated by the ratio MNase-digested versus undigested DNA, remains unaltered in presence of GSI. For B–D, the data points are mean ± SD for triplicate samples from representative experiments.

KDM5A controls expression of Notch target genes. In transient transfection assays, KDM5A acts as a corepressor of RBP-J. Portions (2 μg) of human Hes-1-specific reporter construct (HES-1-Luc) (A) and reporter construct pGA981/6 (B) were transfected alone or cotransfected with 100 ng of plasmids expressing RBP-J-VP16 together with Flag-KDM5A, HA-KDM5A, or catalytic-inactive HA-KDM5A(H483A). Wild-type but not catalytic-inactive KDM5A was able to repress transcription activity. (Y-axis) Relative promoter activity. Data are mean ± SD of three independent experiments ([**] P < 0.01). (C) Nuclear localization of both KDM5A (wt) (top panel) and the catalytic-dead KDM5A (H483A) (bottom panel) after transient transfection into HeLa cells. HA-tagged KDM5A proteins were analyzed by confocal microscopy 24 h after transfection. Bar, 10 μm. (D) Knockdown of KDM5A leads to an increase in expression of Deltex-1 (mean ± SD, n = 5; [*] P < 0.05, Student's t-test). (E) In KDM5A knockdown cells, H3K4 trimethylation is increased at the RBP-J-binding site (−1.2 kb) of the Deltex-1 promoter, but not at the TSS. The values are mean ± SD for triplicate samples from a representative experiment ([*] P < 0.05, Student's t-test). (F) Knockdown of RBP-J leads to an up-regulation of Deltex-1 expression (mean ± SD, n = 3; [*] P < 0.05, Student's t-test). (G) In ChIP experiments using α-KDM5A antibodies, the recruitment of KDM5A to the Deltex-1 promoter is abrogated in RBP-J knockdown cells. At a Notch-independent gene, mitochondrial TOM22, KDM5A localization remains constant in the presence or absence of RBP-J. The values are mean ± SD for triplicate samples from a representative experiment ([*] P < 0.05, Student's t-test).

Model for dynamic H3K4 trimethylation at Notch target genes: When Notch signaling is active, Notch^ICD migrates into the nucleus, associates with transcription factor RBP-J, and recruits Mastermind (MAML) and other coactivators (CoA) including a histone acetyltransferase (HAT). Upon GSI treatment, Notch^ICD gets degraded and RBP-J recruits KDM5A, which demethylates H3K4me3. When GSI is washed off, Notch^ICD replaces KDM5A and recruits a coactivator complex that re-establishes H3K4 trimethylation and, subsequently, transcriptional activity.

Transcriptional regulation of Notch target genes after GSI treatment. (A) cDNA was synthesized from Beko cells before and 24 h after treatment with GSI (DAPT). A miniarray of Notch-relevant genes shows that only very few genes (Deltex-1, Hes-1, preTα, and CD25) are down-regulated upon GSI treatment (red, in blue housekeeping genes). The gray lines indicate a twofold change. (B) Expression of Deltex-1, Hes-1, and preTα is down-regulated rapidly after GSI treatment, as measured by RT–PCR comparing GSI versus control (DMSO-treated) cells (mean ± SD, n = 3; [**] P < 0.01, Student's t-test). (C) In ChIP experiments, H3K4 trimethylation at the conserved RBP-J-binding site (−1.2-kb upstream of the TSS) (see Fig. 2) at the Deltex-1 promoter disappears after GSI treatment with a half-life of 4 h. In the mock control, only the beads were used. Values are mean ± SD for triplicate samples from a representative experiment. (D) The transcriptional down-regulation by GSI treatment is reversible: Washing away GSI after 24 h leads to a rapid up-regulation of Deltex-1, Hes-1, preTα, and CD25. (Mean ± SD, n = 3, [**] P < 0.01, Student's t-test.)

RBP-J and histone demethylase KDM5A physically interact in vitro and in vivo and colocalize at promoters. (A) RBP-J and KDM5A interact in GST pull-down experiments: Cell-free synthesized ³⁵S-labeled KDM5A binds to GST-RBP-J immobilized to glutathione-Sepharose beads, but not with GST only. Cell-free synthesized ³⁵S-labeled KDM5C binds to neither GST-RBP-J nor GST only. (B,C) RBP-J and KDM5A interact in coimmunoprecipitation experiments: HEK293 cells were cotransfected with expression plasmids for Flag-RBP-J, together with Flag/HA-tagged KDM5A; see expression and input controls in B. (B) The expression was detected by Western blotting (WB) using antibodies against the Flag tag. (C) RBP-J was coimmunoprecipitated with only KDM5A from lysates of cells cotransfected together with HA-KDM5A. Coimmunoprecipitated RBP-J was detected using an α-RBP-J antibody. (D–F) KDM5A is part of an endogenous RBP-J complex in Beko T cells: Purification of endogenous RBP-J complexes from Beko cells by DNA affinity chromatography. For purification, a DNA fragment containing 12 RBP-J-binding sites was biotinylated and immobilized on a streptavidin-Sepharose column. The column was incubated with whole-cell lysates from Beko cells. After incubation and washing, the DNA-bound RBP-J complexes were eluted with increasing concentrations of NaCl and were analyzed by EMSA (D). Western blot analysis (WB) of the lysate, the first wash step, and the eluted fractions (E1–E4) were performed using antibodies against RBP-J (E) and KDM5A (F, top panel) and control NF-κB (F, bottom panel). KDM5A, but not p65, is detected in the elution step E2 together with RBP-J. (G) Mapping of the KDM5A-interacting domain: Cell-free synthesized ³⁵S-labeled RBP-J was incubated with different bacterially expressed GST fusion proteins of KDM5A (JmjN/ARID, PHD1, JmjC/Zn finger, PHD2, Inter, or PHD3) (see also Supplemental Fig. S6A,B). (Lanes 7,8) The Inter region and the PHD3 domain containing the C-terminal part of KDM5A interact with RBP-J. (H) Mapping of the RBP-J-interacting domain: Different cell-free synthesized ³⁵S-labeled RBP-J constructs (RBP-J-NTD, RBP-J-β-treefoil [BTD], and RBP-J C-terminal [CTD]) (see Supplemental Fig. S6C) were incubated with bacterially expressed and immobilized GST-Inter or GST-PHD3 fusion proteins. The NTD and BTD of RBP-J interacts with the Inter-PHD2/3 of KDM5A, whereas the PHD3 of KDM5A binds to only the NTD of RBP-J. (I) ChIP analysis using an α-RBP-J antibody (black bars) or mock (IgG control, white bars) at the Deltex-1 promoter of control GAPDH promoter. As described previously (Kathrein et al. 2008), RBP-J is found at −1.2 kb. Data are mean ± SD of two independent experiments measured in triplicate ([*] P < 0.05, Student's t-test). (J) ChIP analysis with an α-KDM5A antibody (black bars) or mock (no antibody, white bars). KDM5A was found at the RBP-J-binding site (1.2 kb), but also at −1.5 kb and at −900 bp. Analysis was performed 8 h after GSI treatment (see also K). No occupancy was found at the GAPDH promoter. Values are mean ± SD for triplicate samples from a representative experiment ([*] P < 0.05, Student's t-test). (K) ChIP time-course experiment with α-KDM5A-specific antibody at 0, 2, 4, 6, 8, and 24 h after GSI treatment. KDM5A occupancy at the RBP-J-binding site increases upon GSI treatment and reaches its maximum after 8 h. Values are mean ± SD for triplicate samples from a representative experiment and were normalized to the levels of KDM5A at Notch-independent KDM5A target TOM22, measured with the same samples ([*] P < 0.05, Student's t-test) (see Fig. 4G).

In vivo interactions of Lid and Su(H) in growth control and tumorigenesis in Drosophila. (A–C) Drosophila RBP-J, called Su(H), and Drosophila KDM5A, called Lid, interact in GST pull-down experiments. (A) Positive control Hairless and Lid were synthesized and ³⁵S-labeled. Hairless (H in B) and Lid (C) bind to GST-Su(H) (lanes 6,8) and to GST-RBP-J (lanes 4,10) immobilized to glutathione-Sepharose beads. (Lanes 3,5,7,9) Lid or Hairless do not interact with GST alone. (D–G) The size of female adult eyes from identically reared animals were scored. (D) Wild type (wt). (E) Overexpression of fringe by ey-Gal4 (abbreviated ey-Gal4 > fng), a Notch pathway modulator, results in a small eye phenotype (∼12% of wild-type eye size). (F,G) Reducing one gene copy of Su(H) (Su[H]^del47/+) (F) or lid (lid^10424/+) (G) consistently rescued the growth defect caused by reducing Notch pathway activation (60%–70% of wild-type eye size; penetrance 100%). (H) Quantification of adult fly eye tumors and metastasis associated with Notch ligand Delta gain of function and the epigenetic repressors Pipsqueak and Lola (eyeful) control cross and the phenotype in the presence of a mutant copy of lid (+ lid¹⁰⁴²⁴/+ +), Su(H), and lid (Su[H]² lid¹⁰⁴²⁴/+ +) or Su(H) mutation alone (Su[H]² +/+ +). Bars shown represent the mean of total (n > 100) flies scored. Crosses were repeated twice. The overexpression of Delta and pipsqueak and lola genes (eyeful) in the developing eye results in eye overgrowth (25.10% of animals, with >130% increased wild-type eye size) and massive eye tumor overgrowth (65.8% of animals), and metastases in distant tissues within the thorax (9.2% of mutant flies, n > 100). Reducing lid or Su(H) increased the frequency of animals with eye tumor growth and metastases. Reducing lid using a different mutation and RNAi lead to similar enhancement of tumorigenesis (see Supplemental Fig. S9E). (I) Wild-type (wt)wing. (J) Su(H)2 lid10424/+lid10424) wing. Arrow points to vein loss in the male wing when a mutant copy of Su(H) is introduced in a lid mutant background.